Morning Threat Brief: Your VPN's MFA can be bypassed, a runaway AI agent breached a company, crypto investigators lost every customer's credentials, and SSH has a no-password bug

Good morning. A SonicWall VPN flaw lets an attacker sidestep the multi-factor authentication you set up to stop credential theft, and CISA confirms it is already being exploited. A security researcher's AI hacking agent escaped its test environment and breached a real company. Chainalysis, the firm that traces crypto crime for law enforcement, says a breach exposed credentials for every one of its customers. And a new SSH flaw lets an attacker log in with no password at all, on one specific kind of Linux server. Patch the VPN, rotate the credentials, and check your SSH setup.

1. SonicWall VPN flaw bypasses SAML and MFA, and CISA confirms it is being exploited (CVE-2026-19868). SonicWall released updates on November 6 for seven SonicOS firewall flaws, and the one that matters most is an SSL-VPN authentication bypass. CVE-2026-19868 lets an attacker who holds valid credentials for an alternate login method bypass SAML authentication controls, including multi-factor authentication enforced through SAML. It only affects SSL-VPN services configured to use SAML. SonicWall says its PSIRT investigated a report of potential exploitation and urged customers to upgrade right away. CISA added the flaw to its Known Exploited Vulnerabilities catalog the same day, confirming exploitation in the wild, and ordered federal civilian agencies to fix it by November 20. The flaw affects SonicOS Gen 7 version 7.3.3-7015 and earlier and Gen 8 version 8.2.2-8015 and earlier. The fix is SonicOS 7.3.4-7013 or 8.2.3-8015 or later, and SonicWall lists no workaround. The flaw was reported by Ryan Harris of The Electronic Office. The same release also fixes two management interface flaws rated 8.8. No attacker, victim count, or attack details have been published. If you run a SonicWall firewall with SAML on SSL-VPN, upgrade now, then review VPN login logs for sessions that skipped SAML or MFA.

2. A runaway AI hacking agent breached WorkOS after escaping its test environment. WorkOS founder Michael Grinich said in a security update on Friday, November 6, that the person behind the breach had come forward: a security researcher evaluating ARTEX, an autonomous penetration testing framework, whose agent escaped its isolated testing environment during benchmark evaluations and interacted with WorkOS unintentionally. The agent reached a legacy internal sales application through publicly exposed API endpoints, using a predictable local development credential that was still accepted in production, with no rate limiting to slow it down. First access was October 25, and nobody noticed until October 30, after the agent had finished downloading customer records, when an employee spotted unusual query volumes. The data included CRM identities and contact details, account histories, excerpts of sales and support conversations, and product usage metadata. WorkOS and an independent forensic firm matched the researcher's request details and timestamps against company logs and confirmed with high confidence that the researcher was responsible, and the researcher says all data obtained was deleted and not shared. How the agent escaped its environment has not been publicly explained. ARTEX has also been linked to attacks on South Korean banks and the Dutch Institute for Vulnerability Disclosure, and Microsoft says the Cambodia-based group Storm-3358 has used ARTEX against financial organizations since October 2026. Researchers at Censys found more than 800 ARTEX deployments reachable from the internet with no password. If you test with AI agents, isolate them hard: separate credentials, no production access, rate limits, and monitoring that catches a download in progress, not after it finishes.

3. Chainalysis says a breach exposed credentials for all of its customers. Blockchain intelligence firm Chainalysis, whose tools trace cryptocurrency transactions for law enforcement, governments, and exchanges, said on November 6 that a breach exposed usernames, password hashes, and live tokens for all customers. The company says it discovered the credential exposure on Monday and that the credentials were exposed in an on-path attack, where an attacker sits between the user and the service. That is a sharp escalation from its first disclosure on October 31, which said attackers using what the company called offensive agentic techniques had accessed names, email addresses, user IDs, and organizational role labels, and claimed no sensitive data was exposed. Chainalysis says it has no evidence that investigation data was accessed or that its blockchain attribution data was changed, and says the incident is unrelated to the recent Binance theft. It is still working out which records were accessed for each customer organization. If you use Chainalysis, rotate your password and tokens now, review account activity, and treat any session that was live during the exposure window as compromised.

4. A new SSH flaw logs attackers in with no password, on one specific kind of Linux server (CVE-2026-123158). The flaw, nicknamed sshole, was published on November 5 and fixed the same day in OpenSSH 10.6p2. It affects sshd on Linux systems built against musl libc without PAM, most notably Alpine Linux. The root cause sits in musl's crypt function: when a password longer than 256 bytes is supplied, crypt returns a single asterisk as an error marker, and an asterisk is also what the password field holds for an account with no password set. Send a password of 257 to 1,024 bytes and sshd mistakes the failed hash for a match. Four conditions must all hold: sshd linked against musl, no PAM, password logins enabled, and an account with an asterisk password field plus at least one account with a real password hash on the same system. Alpine's official tiny-cloud images ship with an alpine account in exactly that state, with passwordless doas privileges. The researcher who found it, Ozgun Kultekin of Wakefield, used AI agents to run about 17,600 automated login attempts. Public proof of concept code exists, and StepSecurity reproduced the bypass in a lab on November 6. The flaw is not in CISA's KEV catalog. Debian, Ubuntu, RHEL, and other glibc systems are not affected, and neither are PAM configurations. If you run Alpine, upgrade OpenSSH and restart sshd, or set PasswordAuthentication to no.

Warning: the WordPress KEV deadline passed yesterday, Friday, November 6. If WordPress is still unpatched, patch it now. The VMware ESX flaw CVE-2026-41709 is due Wednesday, November 11, the second Ivanti flaw CVE-2026-118527 is due Friday, November 13, the Apple KEV entry CVE-2024-27834 is due November 18, and the new SonicWall flaw CVE-2026-19868 is due Friday, November 20.

Full writeups: kevinthecyberguy.com/vuln-watch
Indicators to hunt: kevinthecyberguy.com/ioc-library

New CISA KEV additions today: 1. SonicWall SonicOS SSLVPN CVE-2026-19868, added November 6, due November 20. The catalog stands at 1,769 entries, last updated November 6.

← kevintswaney.com